Do you actually need an AI agency?
I’ve been thinking about this because there are obviously a lot of new companies appearing with AI in the name, and plenty of established companies repositioning themselves around it too. I completely understand why. Businesses want to use AI, so naturally they go looking for an AI company. But I’m not sure we’ve treated many other technologies this way.
If a client needed Stripe payments five years ago, we didn’t suddenly become a payments agency. When we started building more things in the cloud, we didn’t become a cloud agency. We’ve built mapping systems, integrations, apps, ecommerce platforms and all sorts of other things over the years without deciding that one particular piece of technology defined what kind of company we were. We learned how to use it and, when it was useful, used it.
AI is obviously a much bigger shift than adding Stripe to a website. It’s changing what we can build, how quickly we can build it and increasingly how we actually do the building. There are things we’re talking about with clients now that I wouldn’t have considered realistically achievable a few years ago, and I’m genuinely excited about where that’s going.
What I find interesting though is that the problems clients bring us haven’t really changed. Nobody actually wants a large language model for the sake of owning one. They want to stop somebody spending half a day processing something manually, make sense of years of documents, get information from one system into another, or replace the spreadsheet that’s somehow ended up running half the company.
AI gives us some very different ways of solving those problems, but it’s still part of a bigger piece of software. It needs to work with the systems already there, use the right data, respect permissions, handle the weird exceptions and have some thought given to what happens when it gets something wrong. Quite a lot of that isn’t really AI work at all. It’s the same sort of software development work we’ve been doing for 13 years.
Sometimes AI might be almost the whole solution. Sometimes it might be one clever bit in the middle of an otherwise fairly conventional system. Sometimes we’ll look at the problem and decide that AI adds cost and uncertainty to something we could solve perfectly well without it.
There are absolutely projects where a specialist AI company is the right choice. There are people doing serious work with machine learning and developing models at a level we don’t pretend to operate at. But I don’t think that’s what most SMEs mean when they say they want to “use AI”. Usually they’ve seen what is now possible and they’re wondering whether some horrible bit of their business could work better.
That’s the bit I think we’ve slightly got backwards. Rather than deciding you need AI and then finding somebody to sell it to you, show a good software company the problem and see what they’d build.
AI might turn out to be a huge part of the answer. It increasingly is for us. But I don’t think we need to put it in our name.